Fatigue, physical activity and associated factors in 779 patients with myasthenia gravis

The objective of the study was to examine the association between fatigue (measured by the Multidimensional Fatigue Inventory; MFI-20) and physical activity (measured by the Saltin-Grimby Physical Activity Level Scale; SGPALS) in a large cohort of patients (≥18 years) with myasthenia gravis (MG) including relevant disease - and lifestyle-related factors. A total of 1463 persons, registered at the Danish National Registry of Patients with a MG diagnosis, according to the International Classification of Diseases, received a web-based survey. A total of 779 patients (53% women, mean [SD] age 60.8 [15.5]) responded. The remaining persons were either non-responders (n = 390) or could not confirm the MG diagnosis (n = 294). The most prominent MFI-20 fatigue domains were general fatigue (median [inter-quartile ranges, IQR], 13 [10-16]) and physical fatigue (median [IQR], 13 [9-15]), and 386 (53%) patients reported low levels of physical activity. All fatigue domains were associated with physical activity (p<.01). Higher level of physical activity was associated with lower levels of fatigue. Important factors for the association were myasthenia gravis disease severity (measured by the Myasthenia Gravis Activities of Daily Living profile), body mass index, insomnia (measured by the Insomnia Severity Index) job-status, comorbidity, and cohabitation.


Introduction
Myasthenia gravis (MG) is a chronic, autoimmune disease with fluctuating strength of voluntary muscles and distinct muscular fatigue. However, many patients with MG in our clinic also complain about another type of fatigue; a severe, overwhelming and constant fatigue that does not disappear at rest and is perceived different from the well-known and well-described muscular fatigue.
No standard definition of fatigue exists, although many suggestions have been proposed. In one attempt of defining only cover one or two fatigue domains, e.g., cognitive, mental and/or physical fatigue, summarize one total score across fatigue domains, include items not relevant for MG patients, or are to comprehensive to include in a survey. Using a multidimensional assessment tool for future research on fatigue in MG has been recommended in a recent systematic review of fatigue in MG [2] . The Multidimensional Fatigue inventory (MFI-20) [9] is such an assessment tool, which is easy to use and includes five fatigue domains; physical fatigue, general fatigue, reduced activity, reduced motivation and mental fatigue. This tool and these fatigue terms are used in the present study, where the domain, physical fatigue, is considered to correspond to the definition of peripheral fatigue, whereas the remaining domains represent different aspects of the central fatigue.
Recent trials [ 10 , 11 ], as well as older case-studies of patients with MG participating in endurance [12][13][14] or resistance training [ 15 , 16 ], suggest that exercise training is beneficial, especially by increasing muscle strength without apparent side effects. However, only one small study [17] has examined the level of habitual physical activity in patients with MG. Habitual physical activity refers to the physical activity (PA) that is integrated in peoples' everyday life in their natural environment, and thus covers the activity level of a whole day instead of dedicated exercise sessions. Habitual physical activity is a relevant and important outcome measure in research, as lifelong habitual PA at recommended intensities is important to obtain the well-known health benefits of being physically active. No previous studies have examined the influence of fatigue on habitual physical activity levels in MG, and a better understanding of the potential association between different types of fatigue and PA in these patients is needed.
The objective of the study was to describe fatigue and PA levels in a national cohort of patients with MG, and to examine the association between different fatigue domains and PA levels, including relevant disease -and lifestyle-related factors.

Materials and methods
All Danish citizens are assigned a unique 10-digit civil registration number at birth or upon immigration. This has been used since 1977 to store data on all hospital inand outpatient discharges in the Danish National Registry of Patients (DNRP). Most Danes use e-Boks, a personal, digital mailbox, connected to the civil registration number, to communicate sensitive information e.g. with health authorities.
In this study, we obtained permission to extract contact information from the Danish National Registry of Patients on persons registered with a diagnostic code of MG. The persons were ≥18 years, residents of Denmark, subscribed to e-Boks, and were coded according to International Classification of Diseases eight edition (ICD-8: 733.09, 1971(ICD-8: 733.09, -1993 or tenth edition (ICD-10; G.70.0, from 1994) in the period from 1977 to end of 2018.
Persons, who met these criteria, were invited to participate in a web-based survey, using the software REDCap (© 2018 Vanderbilt University). All responses to the survey were automatically stored in a secure database. The invitation was sent out three times; first, in June 2019 and to non-responders in July 2019 and November 2019.
The study was approved by the ethics committee of the Capital Region of Denmark (approval H-18,031,231), registered at the Danish Data Protection Agency (VD-2018-440, I-Suite nr.6694) and approved by the Danish Health Data Authority. Written informed consent was obtained from all participants.

Fatigue
The MFI-20 [9] is a generic, self-reported questionnaire that measures fatigue severity. The development of the questionnaire was based on comprehensive patient interviews, literature studies and theoretical considerations [18] . In the original validation study, Cronbach's alpha was reported to be 0.84 [9] . The MFI-20 has been validated and/or used in several clinical and healthy populations since 1995, including patients with neurological diseases, e.g. multiple sclerosis, stroke, polio and spinal muscular atrophy [19][20][21][22][23][24][25] . The MFI-20 was in 2000, in a study of fatigue in the Danish background population, validated and translated into Danish [ 26 , 27 ]. MFI-20 contains 20 items and categorises fatigue into five domains: general fatigue (e.g., "I feel tired"), physical fatigue (e.g., "Physically, I feel I am in bad condition"), reduced activity (e.g., "I get little done"), reduced motivation (e.g., "I don't feel like doing anything") and mental fatigue (e.g., "It takes a lot effort to concentrate on things"). The response options consist of five check boxes ranging from "Yes, that is true" to "No, that is not true". The total score in each domain ranges from 4 to 20, with higher scores indicating higher levels of fatigue. A total score across fatigue domains is not recommended.

Physical activity level
The Saltin-Grimby Physical Activity Level Scale (SGPALS), which was developed in 1968 [28] , has been modified in various ways [ 29 , 30 ]. As the SGPALS is an easy-to-use measurement tool for self-reported PA, it has been used in several large-scale population-based studies, also in Denmark [31][32][33][34][35] . SGPALS has been used in patients with stroke [36][37][38][39] , but never in patients with MG. High reliability and validity of SGPALS has been demonstrated [40] . In the present study, we measured PA levels based on the original questionnaire, but with a slightly modified wording both in the following question and in the four response categories: "During the last year, how physically active have you been, including leisure time activities and transportation? Please categorize yourself into one of four levels of physical activity": I (Sedentary). Mainly sedentary or engaged in light physical activity less than 2 h per week (e.g., reading books, watching television, or going to the cinema). II (Light to moderate). Light-to-moderate physical activity 2-4 h per week (e.g., walking, cycling for pleasure, gardening, housework, light exercise). III (Regular moderate). Moderate physical activity more than 4 h per week, or more strenuous activities 2-4 h per week (e.g. brisk walking, fast bicycling, heavy gardening, or exercises that makes you short of breath). IV (Regular vigorous). More strenuous physical activities more than 4 h per week, or regular vigorous exercise (e.g. competitive sport) several times a week.

MG severity
The Myasthenia Gravis Activities of Daily Living profile (MG-ADL) is an 8-item patient-reported questionnaire where a higher score indicates higher MG severity (total score range 0-24) [41] . The scale assesses common MG-symptoms and dysfunctions, including questions of ocular, bulbar, respiratory, and extremity functions. Patients were asked about symptoms endured in the past seven days. The MG-ADL is found valid as an outcome measure in research and clinical practice [42] .

Quality of life
The MG-specific quality-of-life instrument (MG-QoL15) is a 15-item questionnaire assessing quality of life in patients with MG. Rating consists of a 5-point scale ranging from 0 to 4 indicating the patient's agreement with a given statement; summing up to a total score of 0-60 points (60 = low quality of life). The MG-QoL15 was developed in 2008 [43] , and has been found valid and reliable [ 8 , 44 , 45 ] in patients with MG. The MG-QoL15 has been revised (MG-QoL15r) [46] in 2016, but for comparison of results from older studies, the original MG-QoL15 is used in present study.

Insomnia
The Insomnia Severity Index (ISI) (Mapi Research Trust, Lyon, France) measures self-reported sleep quality and disturbances during the previous 2 weeks. The ISI comprises seven items assessing the perceived severity of insomnia, each item rated on a five-point scale (total score 0-28, most severe insomnia = 28). A total score ≥ 8 indicate clinical insomnia. ISI was developed in 1993 [47] , used in several clinical and healthy populations, including patients with MG [3] . The reliability and validity of ISI has been tested and found to be good [48] .

Background information
Patients reported age, sex, height + weight (used to calculate body mass index, BMI), MG duration, current job status, frequency of follow-up at neurologist, thymectomy and family status (cohabitant with person(s) ≥ 18 years). In addition, there were questions (with yes/no answers) about comorbidities relevant for fatigue and physical activity, and questions about the most common medical treatment for MG and comorbidity.

Statistical analyses
Continuous variables were presented by means and standard deviation (SD). Non-parametric continuous variables were presented by medians and inter-quartile ranges (IQR). Normality was assessed visually by histograms and boxplots. Categoric variables were presented by numbers and percentages.
For a descriptive purpose, the SGPALS was categorized into two groups; low PA (levels I and II) and regular PA (levels III and IV). Differences in characteristics between low and regular levels of PA were investigated by unpaired t-test for continuous data, by Mann-Whitney test for non-normal continuous data, and by Fisher's exact test for categoric data, and presented in Table 2 . We did not present such a descriptive table for fatigue, as no cut-off points for low/high fatigue exist for MFI-20. However, differences were reported in text, based on the sample median score in each fatigue domain.
To examine the association between fatigue and PA, a general linear regression model was applied, using MFI-20 scores in each fatigue domain as continuous outcome variables. Covariates, included in the statistical models, were selected a priori, and determined from known evidence, or experiences from the clinic. Age, BMI, MG duration, MG-ADL, MFI-20 and ISI scores were included as continuous variables. Sex, job-status, cohabitation, medication other than MG treatment, and comorbidity were included as categoric variables.
Collinearity among covariates were examined by correlation analyses included in the models. To prevent collinearity, MG-QoL15 was not included in the statistical models as collinearity was expected between MG-QoL15 and MG-ADL because many of the items in these scales are alike. The regression models were executed as complete-case analysis and checked by 1) goodness of fit test, 2) test for linearity of covariates by adding log transformed covariates into the model, 3) test of accumulated residuals by plots and p-values. Convergence criterion was satisfied in all analyses.
Some responders confirmed their MG diagnosis but were not regularly followed by a neurologist and/or pharmacologically treated for MG. Therefore, to avoid misleading conclusions due to including potential non-MG persons, all analyses were additionally executed on a subsample of patients ( n = 486). These patients were regularly followed by a neurologist and were on active MG treatment.
For analyses, a p ≤ 0.05 (2-tailed testing) was considered significant. All statistical analyses were carried out using SAS enterprise guide 7.1. MG according to the  ICD 8 or ICD 10 from 1977 to end 2018,  n=1745 PaƟents excluded due to:

Danish residents coded for
Not subscribing e-Boks, n=282 Not recognizing MG diagnosis, n=294 Not-responding, n=390 Responders n=779

Study population
A total of 1745 persons were registered at the DNPR with a diagnostic code of MG. However, 282 persons (mean age [SD], 78 [12.6] years, women 164 [58%]), did not subscribe to e-Boks, and 390 persons (mean age [SD], 55 [16.5] years, women 226 [58%]) never responded to the survey and were excluded. Another 294 persons phoned or emailed the study coordinator (LKA) because; 1) they were mistakenly registered at the DNPR with a MG diagnosis, but suffered from another neuromuscular disease, or 2) they haven't had any MG symptoms for years and were uncertain of the MG diagnosis. These 294 patients were excluded. This left 779 participants who completed the survey either completely ( n = 662) or partially ( n = 117). Of these 779 patients, 499 (64%) responded the first time the invitation was sent out (June 2019), 193 (25%) the second time (July 2019) and 87 (11%) the last time (November 2019).
The response rate was calculated according to recommended standards from the American Association for Public Opinion Research [49] . Response rate = (662 + 117/ (662 + 117) + e (282 + 390)), where e is the estimated proportion of persons of unknown MG status that were expected to have a MG diagnosis. As the Danish MG population is estimated to be 1000 patients [50] , and as 779 patients already responded, we expected that 221 (approximately one third) of the 672 with unknown eligibility ( e = 0.33) likely could have a diagnosis of MG. Therefore, the response rate of the survey was set at 78%. Enrolment pathway is illustrated in Fig. 1 .
The demographic characteristics of the patient samples are presented in Table 1 . The distribution of characteristics in the sub-sample ( n = 486) was very similar to the overall sample.

Fatigue
The MFI-20 domains general fatigue (median 13, IQR 10-16), physical fatigue (median 13, IQR 9-15) and reduced activity (median 12, IQR 9-15) were the most pronounced fatigue domains in this cohort ( Table 1 ). Reduced motivation and mental fatigue were less pronounced with a sample median score of 9. This distribution was similar in the sub-sample ( n = 486), and in patients not regularly followed by a neurologist and not on MG treatment ( n = 144). The distribution pattern was similar in patients reporting comorbidity/no comorbidity, even though comorbid patients had higher scores in all fatigue domains vs. patients reporting no comorbidity ( p < .01). Also, patients with another autoimmune disease besides MG and patients treated with steroids had higher fatigues scores than the remaining MG patients ( p < .01).

Physical activity
The distribution of SGPALS levels ( Table 1 ), resulted in 386 (53%) patients categorized in the group with low PA Legend Table 1    (levels I + II) and 345 (47%) categorized in the group with regular PA (levels III + IV). An almost similar distribution was seen for the sub-sample (low PA: 50%, regular PA: 50%). Compared to patients in the group with regular PA ( Table 2 ), patients in the group with low PA were slightly older (mean age 62 vs 59, p < .01), had a higher mean BMI (27.9 vs 26.4, p < .01), more often reported comorbidity (69 vs 58%, p < .01), more MG symptoms (MG-ADL, median 3 vs 2, p < .01), lower QoL (MG-QoL15, 12 vs 7, p < .01), more sleep problems (ISI, 9 vs 7, p < .01), and reported higher levels of fatigue in all domains ( p < .01). These patterns were also present in the subsample.

Fatigue and physical activity
The association of fatigue and PA in the overall sample and the sub-sample are presented in Tables 3 and 4 , respectively. There was a strong significant association between the MFI-20 fatigue domains and SGPALS, indicating that higher level of PA was associated with lower levels of fatigue, in both the overall sample and in the sub-sample, except for mental fatigue in the sub-sample. Statistically significant covariates influencing on this association were primarily MG severity (MG-ADL) and insomnia (ISI). Also, BMI, comorbidity (e.g., depression/anxiety), and job-status influenced on the association in the overall sample, but not in the sub-sample, while cohabitation was a statistically significant covariate in the sub-sample.

Discussion
In this survey of 779 patients with MG, self-reported high levels of fatigue and low levels of PA were frequent, and strongly associated. Important covariates in this association were MG disease severity, sleep quality, BMI, comorbidity, job-status, and cohabitation.
The level of general fatigue and physical fatigue were similar in this cohort, and these fatigue domains were the two most pronounced domains in the sample. This was also seen in the Danish background population [27] . Also, the MFI-20 domain for reduced activity was pronounced in this sample, whereas the median sample score of the domains reduced motivation and mental fatigue were lower. In one study of 4964 Danish adults in late mid-life (age 49-63) [51] , the median sample score was 9 for general fatigue and physical fatigue, and 7 for the remaining domains, indicating an increased level of fatigue in patients with MG compared to a healthy population. An increased level of fatigue in MG compared to healthy controls has previously been reported [ 4 , 5 ]. However, patients reporting full remission (MG-ADL score = 0) ( n = 142) had fatigue scores very similar to the non-MG background population [51] . The fatigue scores in the present cohort were slightly lower than found in patients with Parkinson disease [52] , chronic fatigue syndrome [53] , multiple sclerosis [24] and almost similar to fatigue scores in post-stroke patients [20] and patients with other autoimmune diseases [54] . One previous study has published results of MFI measurements in patients with neuromuscular diseases, finding a fatigue level in adult patients with spinal muscular atrophy lower (except for physical fatigue) than found in patients with MG [25] . Due to a large variation in measurement tools of fatigue across studies, comparison of results in previous MG studies are difficult. However, this study was the first to measure the level of fatigue in five different fatigue domains.
The grouping of SGPALS scores into low-and regular PA, as performed in other patient studies [54] , gave us the opportunity to differentiate between patients meeting/not meeting the existing recommendations from WHO [55] . We assumed that patients who reported SGPALS scores III and IV met the WHO recommendations (minimum 150 min per week of moderate intensity), corresponding to 47% of the cohort. The 53% of the patients, not meeting the recommendations, corresponded to findings in patients Table 3 Association of fatigue and physical activity level in the overall sample.   Table 1 . f yes/no, medication other than MG medication e.g., benzodiazepines. Abbreviations: BMI (Body mass index), MG-ADL (MG Activities of Daily Living profile), ISI (Insomnia Severity Index), SGPALS (Saltin-Grimby Physical Activity Level Scale), MFI-20 (Multidimensional Fatigue Inventory). The analyses were made as complete-case analysis ( n = 690). Table 4 Association of fatigue and physical activity level in the sub-sample.  with myotonic dystrophy type 1 [56] , whereas patients with MG seemed more physically active than patients with mitochondrial disease [57] . Comparing our findings of PA levels with findings in the Danish general population did not lead to any firm conclusions. The results from the largescale Danish population studies were heterogeneous, and depending on which study we compared to, patients with MG were either more active or less active than the general Danish population [58][59][60] . The question used in these largescale Danish population studies are found to be strongly dependent on the context of the examination and the mode of the administration of the questionnaire [61] . However, this hampers the possibility to compare the PA levels in MG with healthy counterparts, which is a limitation.
In general, it is noteworthy that assessment of PA level in both clinical and healthy cohorts depends on the chosen outcome measurement. This was the case in a study of PA patterns in 27 patients with MG [17] , where 78% of the patients achieved the recommended minimum average of 64 MET min/day. However, when the study examined other outcome measures, e.g. steps per day, the patients' day was dominated by sedentary behaviour. Comparisons with other study results and recommendations are useful to obtain an overall interpretation of the activity level in the cohort, but larger studies including objective measurements is needed to fully examine the pattern of habitual PA in patients with MG.
Engaging in more PA was associated with less fatigue in both the overall sample and in the sub-sample, especially in the fatigue domains; physical fatigue and reduced activity. This finding could indicate that active patients experienced lower physical fatigue, or less fatigued patients engaged in more PA. More research is needed to determine the direction of the association, which cannot be identified by a cross-sectional study like the present.
Several covariates significantly influenced on the association between fatigue and PA. As reported previously [ 3-6 , 62 , 63 ], an increase in MG-severity was associated with an increased level of fatigue. This was seen in all fatigue domains, in both the overall and the sub-sample. Poor sleep quality (ISI score) strongly influenced on increasing fatigue, which have been reported previously [ 4 , 7 ]. In the present study, 55% of the patients reported insomnia (ISI score ≥8), which agrees with the finding of 59% in another MG cohort [7] . As insomnia is also associated with MG-related quality of life [7] , the sleep problems in MG call for action, and must be taken into consideration when meeting these patients in the clinic. Also, higher BMI was associated with increased physical fatigue. This is interesting as a large proportion of patients (60%) of the cohort reported a BMI ≥ 25 (classified as overweight), which was also found in a previous study [62] . Reasons for obesity can be e.g., inactivity (caused by fatigue) or steroid use. However, the direction of the association cannot be assessed in a cross-sectional study, and prospective studies are needed to unravel this association.
Comorbidity was associated with increased physical fatigue in this cohort. These findings make sense as several diseases are known to induce fatigue in patients. Depression is previous found to be associated with fatigue perception in patients with MG [ 4 , 64 ]. However, in the present study, depression/anxiety only influenced on the fatigue domain reduced motivation, whereas depression was not associated with e.g., physical fatigue or mental fatigue. However, it is important to emphasize that the information on comorbidity (e.g. depression) might be prone to bias as no verification of comorbidity was possible in this study design. However, as the overall rate for depression among patients with MG has previously been found to be around 20% [3], the reported 8% in our study might be an underestimation, as well as the association with fatigue.
Female gender has previously been associated with fatigue in MG [ 4 , 6 , 65 ], which was not found in the present study. Living alone compared to living with another person ≥18 years was associated with an increase in fatigue in the sup-sample. However, this was only seen for the activityoriented fatigue domains; physical fatigue, reduced activity and reduced motivation. For these domains, the largest reduction in fatigue was seen, when patients engaged in PA. An explanation to these findings could be that a cohabitant might motivate the patient for PA, or even participate in MG together with the patient, which may be motivational. The fact, that people living alone reported higher levels of fatigue was also seen in the Danish background population [27] .
Even though potential confounders, included in the analyses, were selected a priori from the literature or from experience in our clinic, the influence of potential residual confounding cannot be ruled out. Interpretation of results should be done with caution knowing that the results indicate associations rather than causality. Self-reported outcomes are prone to bias, e.g. social desirability bias, where responders distort self-reports in a favourable direction. However, this bias has been demonstrated to be less pronounced [66] , especially in an online setting where respondents' perceived privacy is higher [67] .
The MFI-20 is generic, and a new instrument in MG research, even though it has been used and validated in neurological research since 1995. MFI-20 was chosen for its multidimensional approach to fatigue, which suited the objective of this study. The study confirmed our experiences from the clinic that general fatigue had a similar impact on the patient's everyday life as the MG-induced physical fatigue, which indicates that this instrument is useful to measure and separate these two different fatigue domains.
Due to the large sample of MG patients included, our study results are likely to represent the MG population of 1000 patients in Denmark [50] . The excluded persons ( n = 282) that did not subscribe to e-Boks had a higher mean age than the overall sample mean, indicating some selection bias towards the younger population, probably due to the webbased methods of this study. Opposite, the persons that did not respond ( n = 390) were younger compared to the overall sample. The total of excluded persons with unknown MG eligibility resembled the included patients regarding age and sex. However, we cannot rule out that these persons were markedly different from the included patients, regarding other characteristics. The results from the overall sample did not differ from the results in the sub-sample of patients in MG treatment, indicating no selection bias in the overall sample.

Conclusion
In this cross-sectional study, including 779 patients with MG, the levels of fatigue were increased compared to the general population. More than half of the cohort reported low levels of PA. Higher level of PA was associated with lower levels of fatigue, and important factors were MG severity, sleep quality, BMI, comorbidity, job-status and cohabitation. The findings suggest that PA may play an important role in managing fatigue in MG, however, the direction of the association needs further investigation. Also, important nonspecific MG factors, e.g. weight and sleep quality, should be considered when planning rehabilitation programs for patients with MG.

Declaration of Competing Interest
Authors report no conflicts of interest.